[PATCH] glib-2.0: fix ptest failures

[PATCH] glib-2.0: fix ptest failures

[mingli.yu]

From: Mingli Yu <Mingli.Yu@windriver.com>

* Set G_TEST_BUILDDIR to mark we run the tests
  out of builddir to make it can skip the test_non_utf8_printf
  to fix below test error as test_non_utf8_printf needed
  to run under builddir per https://gitlab.gnome.org/GNOME/glib/blob/master/glib/tests/gdatetime.c#L1440.
  | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1477:test_non_utf8_printf: assertion failed (__p == ("10\346\234\210")): ("Oct" == "10\346\234\210")
  | Bail out! GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1477:test_non_utf8_printf: assertion failed (__p == ("10\346\234\210")): ("Oct" == "10\346\234\210")
  | FAIL: glib/gdatetime.test (Child process killed by signal 6)

* Increase the timeout to 15 minute to fix below
  error as the default timeout which is 5 minutes
  is not enough.
  | Executing: glib/actions.test
  | Test timed out after 300 seconds
  | cleaning up pid 13050
  | FAIL: glib/actions.test (Child process killed by signal 9)

Signed-off-by: Mingli Yu <Mingli.Yu@windriver.com>
---
 meta/recipes-core/glib-2.0/glib-2.0/run-ptest | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/meta/recipes-core/glib-2.0/glib-2.0/run-ptest b/meta/recipes-core/glib-2.0/glib-2.0/run-ptest
index 8f082d3..e4777da 100644
--- a/meta/recipes-core/glib-2.0/glib-2.0/run-ptest
+++ b/meta/recipes-core/glib-2.0/glib-2.0/run-ptest
@@ -1,6 +1,8 @@
 #! /bin/sh
 
+# set G_TEST_BUILDDIR to mark the tests running out of the builddir
+export G_TEST_BUILDDIR=$(pwd)
 set -eux
 useradd glib2-test
-su glib2-test -c gnome-desktop-testing-runner glib
+su glib2-test -c "gnome-desktop-testing-runner glib -t 900"
 userdel glib2-test
-- 
2.7.4

Re: [PATCH] glib-2.0: fix ptest failures

[Burton, Ross]

On Fri, 19 Apr 2019 at 09:54, <mingli.yu@windriver.com> wrote:
> From: Mingli Yu <Mingli.Yu@windriver.com>
>
> * Set G_TEST_BUILDDIR to mark we run the tests
>   out of builddir to make it can skip the test_non_utf8_printf
>   to fix below test error as test_non_utf8_printf needed
>   to run under builddir per https://gitlab.gnome.org/GNOME/glib/blob/master/glib/tests/gdatetime.c#L1440.
>   | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1477:test_non_utf8_printf: assertion failed (__p == ("10\346\234\210")): ("Oct" == "10\346\234\210")
>   | Bail out! GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1477:test_non_utf8_printf: assertion failed (__p == ("10\346\234\210")): ("Oct" == "10\346\234\210")
>   | FAIL: glib/gdatetime.test (Child process killed by signal 6)

As per the link:

/* If running uninstalled (G_TEST_BUILDDIR is set), skip this test, since we
   * need the translations to be installed. We can’t mess around with
   * bindtextdomain() here, as the compiled .gmo files in po/ are not in the
   * right installed directory hierarchy to be successfully loaded by
gettext. */

Setting G_TEST_BUILDDIR isn't right as we're not running uninstalled.
This is a gross hack, and not a fix.

Ross

Re: [PATCH] glib-2.0: fix ptest failures

[Yu, Mingli]

On 2019年04月23日 23:52, Burton, Ross wrote:
> On Fri, 19 Apr 2019 at 09:54, <mingli.yu@windriver.com> wrote:
>> From: Mingli Yu <Mingli.Yu@windriver.com>
>>
>> * Set G_TEST_BUILDDIR to mark we run the tests
>>    out of builddir to make it can skip the test_non_utf8_printf
>>    to fix below test error as test_non_utf8_printf needed
>>    to run under builddir per https://gitlab.gnome.org/GNOME/glib/blob/master/glib/tests/gdatetime.c#L1440.
>>    | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1477:test_non_utf8_printf: assertion failed (__p == ("10\346\234\210")): ("Oct" == "10\346\234\210")
>>    | Bail out! GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1477:test_non_utf8_printf: assertion failed (__p == ("10\346\234\210")): ("Oct" == "10\346\234\210")
>>    | FAIL: glib/gdatetime.test (Child process killed by signal 6)
>
> As per the link:
>
> /* If running uninstalled (G_TEST_BUILDDIR is set), skip this test, since we
>     * need the translations to be installed. We can’t mess around with
>     * bindtextdomain() here, as the compiled .gmo files in po/ are not in the
>     * right installed directory hierarchy to be successfully loaded by
> gettext. */
>
> Setting G_TEST_BUILDDIR isn't right as we're not running uninstalled.
> This is a gross hack, and not a fix.

Per the comments at https://gitlab.gnome.org/GNOME/glib/issues/1357, we 
need to run test_non_utf8_printf under builddir to guarantee the test 
passed.

Thanks,

>
> Ross
>

Re: [PATCH] glib-2.0: fix ptest failures

[Burton, Ross]

On Wed, 24 Apr 2019 at 02:51, Yu, Mingli <mingli.yu@windriver.com> wrote:
> Per the comments at https://gitlab.gnome.org/GNOME/glib/issues/1357, we
> need to run test_non_utf8_printf under builddir to guarantee the test
> passed.

Re-read the comments.  Setting that variable skips the test, as it
isn't useful without the translations being installed. So, we need to
install the translations:

root@bob:~# /usr/libexec/installed-tests/glib/gdatetime -p
/GDateTime/non_utf8_printf
/GDateTime/non_utf8_printf: **
GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1477:test_non_utf8_printf:
assertion failed (__p == ("10\346\234\210")): ("Oct" ==
"10\346\234\210")
Aborted (core dumped)

root@bob:~# dnf install glib-2.0-locale-ja
[ ... ]
Installed:
  libglib-2.0-locale-ja-1:2.58.3-r0.corei7_64

root@bob:~# /usr/libexec/installed-tests/glib/gdatetime -p
/GDateTime/non_utf8_printf
/GDateTime/non_utf8_printf: OK

Ross

Re: [PATCH] glib-2.0: fix ptest failures

[Yu, Mingli]

On 2019年04月24日 20:07, Burton, Ross wrote:
> On Wed, 24 Apr 2019 at 02:51, Yu, Mingli <mingli.yu@windriver.com> wrote:
>> Per the comments at https://gitlab.gnome.org/GNOME/glib/issues/1357, we
>> need to run test_non_utf8_printf under builddir to guarantee the test
>> passed.
>
> Re-read the comments.  Setting that variable skips the test, as it
> isn't useful without the translations being installed. So, we need to
> install the translations:
>
> root@bob:~# /usr/libexec/installed-tests/glib/gdatetime -p
> /GDateTime/non_utf8_printf
> /GDateTime/non_utf8_printf: **
> GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1477:test_non_utf8_printf:
> assertion failed (__p == ("10\346\234\210")): ("Oct" ==
> "10\346\234\210")
> Aborted (core dumped)
>
> root@bob:~# dnf install glib-2.0-locale-ja
> [ ... ]
> Installed:
>    libglib-2.0-locale-ja-1:2.58.3-r0.corei7_64

Many thanks for Ross's comments!

Yes, it's true. Install libglib-2.0-locale-ja-2.58.3-r0 did silence the 
test_non_utf8_printf failure of gdatetime.test.

But once unset G_TEST_BUILDDIR, we should check the translation modules 
one by one to silence other failures such as below to guarantee 
gdatetime.test passed.

GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1683:test_month_names: 
assertion failed (p_casefold == (o_casefold)): ("sep" == "sept.")
Bail out! 
GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1683:test_month_names: 
assertion failed (p_casefold == (o_casefold)): ("sep" == "sept.")
FAIL: glib/gdatetime.test (Child process killed by signal 6)

Anyway, I will check the translation issues one by one and resent the patch.

Thanks,

>
> root@bob:~# /usr/libexec/installed-tests/glib/gdatetime -p
> /GDateTime/non_utf8_printf
> /GDateTime/non_utf8_printf: OK
>
> Ross
>

[PATCH v2] glib-2.0: fix ptest failures

[mingli.yu]

From: Mingli Yu <Mingli.Yu@windriver.com>

* Install 7 translation packages to fix 7 ptest
  failures such as:
  | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1477:test_non_utf8_printf: assertion failed (__p == ("10\346\234\210")): ("Oct" == "10\346\234\210")
  | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1716:test_month_names: assertion failed (p_casefold == (o_casefold)): ("nov" == "stu")
  | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1731:test_month_names: assertion failed (p_casefold == (o_casefold)): ("jul" == "liep.")

* Increase the timeout to 15 minutes as the default
  timeout which is 5 minutes is not enough to fix below error:
  | Executing: glib/actions.test
  | Test timed out after 300 seconds
  | cleaning up pid 13050
  | FAIL: glib/actions.test (Child process killed by signal 9)

* There are 240+ ptest cases for glib-2.0 and one
  case is gdatetime.test. And there are 53 sub-cases
  under gdatetime.test.

  Among of the 53 sub-cases of gdatetime.test, there
  is one sub-case named test_month_names.

  For test_month_names case, there are also 64 cases
  against it. Comment 2 cases of test_month_names
  as a workaround to fix the below test_month_names
  failure as the issue also exists in the latest version:
  | GLib:ERROR:../glib-2.60.1/glib/tests
  /gdatetime.c:1702:test_month_names: assertion failed (p_casefold == (o_casefold)): ("\316\261\317\205\316\263" == "\316\261\317\215\316\263")
  | GLib:ERROR:../glib-2.60.1/glib/tests/gdatetime.c:1765:test_month_names: assertion failed (p_casefold == (o_casefold)): ("\320\274\320\260\321\217" == "\320\274\320\260\320\271")

  Have reported the above issues to upstream as
  https://gitlab.gnome.org/GNOME/glib/issues/1759
  and as the failure even exists on the latest version 2.60.1
  until now, so comment it as a workaround until the
  upstream fixs it.

Signed-off-by: Mingli Yu <Mingli.Yu@windriver.com>
---
 ...c-workaround-to-fix-test_month_names-1759.patch | 57 ++++++++++++++++++++++
 meta/recipes-core/glib-2.0/glib-2.0/run-ptest      |  2 +-
 meta/recipes-core/glib-2.0/glib-2.0_2.58.3.bb      |  1 +
 meta/recipes-core/glib-2.0/glib.inc                |  7 +++
 4 files changed, 66 insertions(+), 1 deletion(-)
 create mode 100644 meta/recipes-core/glib-2.0/glib-2.0/0001-gdatetime.c-workaround-to-fix-test_month_names-1759.patch

diff --git a/meta/recipes-core/glib-2.0/glib-2.0/0001-gdatetime.c-workaround-to-fix-test_month_names-1759.patch b/meta/recipes-core/glib-2.0/glib-2.0/0001-gdatetime.c-workaround-to-fix-test_month_names-1759.patch
new file mode 100644
index 0000000..155ae0a
--- /dev/null
+++ b/meta/recipes-core/glib-2.0/glib-2.0/0001-gdatetime.c-workaround-to-fix-test_month_names-1759.patch
@@ -0,0 +1,57 @@
+From c40e4f0d0017c943d5c15bb2e8cfd8becf3e0d84 Mon Sep 17 00:00:00 2001
+From: Mingli Yu <Mingli.Yu@windriver.com>
+Date: Tue, 30 Apr 2019 14:32:29 +0800
+Subject: [PATCH] gdatetime.c: workaround to fix test_month_names[#1759]
+
+There are 240+ ptest cases for glib-2.0 and one
+case is gdatetime.test. And there are 53 sub-cases
+under gdatetime.test.
+
+Among of the 53 sub-cases of gdatetime.test, there
+is one sub-case named test_month_names.
+
+For test_month_names case, there are also 64 cases
+against it. Comment 2 cases of test_month_names
+as a workaround to fix the below test_month_names
+failure as the issue also exists in the latest version:
+GLib:ERROR:../glib-2.60.1/glib/tests
+/gdatetime.c:1702:test_month_names: assertion failed (p_casefold == (o_casefold)): ("\316\261\317\205\316\263" == "\316\261\317\215\316\263")
+
+GLib:ERROR:../glib-2.60.1/glib/tests/gdatetime.c:1765:test_month_names: assertion failed (p_casefold == (o_casefold)): ("\320\274\320\260\321\217" == "\320\274\320\260\320\271")
+
+Have reported the above issues to upstream as
+https://gitlab.gnome.org/GNOME/glib/issues/1759
+and as the failure even exists on the latest version 2.60.1
+until now, so comment it as a workaround until the
+upstream fixs it.
+
+Upstream-Status: Inappropriate [workaround]
+
+Signed-off-by: Mingli Yu <Mingli.Yu@windriver.com>
+---
+ glib/tests/gdatetime.c | 2 --
+ 1 file changed, 2 deletions(-)
+
+diff --git a/glib/tests/gdatetime.c b/glib/tests/gdatetime.c
+index 8ab4681..3f9f0c3 100644
+--- a/glib/tests/gdatetime.c
++++ b/glib/tests/gdatetime.c
+@@ -1699,7 +1699,6 @@ test_month_names (void)
+       TEST_PRINTF_DATE (2018,  5,  1, "%OB", "Μάιος");
+       TEST_PRINTF_DATE (2018,  6,  1, "%OB", "Ιούνιος");
+       TEST_PRINTF_DATE (2018,  7,  1,  "%b", "Ιουλ");
+-      TEST_PRINTF_DATE (2018,  8,  1, "%Ob", "Αύγ");
+     }
+   else
+     g_test_skip ("locale el_GR not available, skipping Greek month names test");
+@@ -1762,7 +1761,6 @@ test_month_names (void)
+       TEST_PRINTF_DATE (2018,  8,  1, "%Ob", "авг");
+       /* This difference is very important in Russian:  */
+       TEST_PRINTF_DATE (2018,  5,  1,  "%b", "мая");
+-      TEST_PRINTF_DATE (2018,  5,  1, "%Ob", "май");
+     }
+   else
+     g_test_skip ("locale ru_RU not available, skipping Russian month names test");
+-- 
+2.7.4
+
diff --git a/meta/recipes-core/glib-2.0/glib-2.0/run-ptest b/meta/recipes-core/glib-2.0/glib-2.0/run-ptest
index 8f082d3..893125d 100644
--- a/meta/recipes-core/glib-2.0/glib-2.0/run-ptest
+++ b/meta/recipes-core/glib-2.0/glib-2.0/run-ptest
@@ -2,5 +2,5 @@
 
 set -eux
 useradd glib2-test
-su glib2-test -c gnome-desktop-testing-runner glib
+su glib2-test -c "gnome-desktop-testing-runner glib -t 900"
 userdel glib2-test
diff --git a/meta/recipes-core/glib-2.0/glib-2.0_2.58.3.bb b/meta/recipes-core/glib-2.0/glib-2.0_2.58.3.bb
index 733a2d4..83fcb69 100644
--- a/meta/recipes-core/glib-2.0/glib-2.0_2.58.3.bb
+++ b/meta/recipes-core/glib-2.0/glib-2.0_2.58.3.bb
@@ -17,6 +17,7 @@ SRC_URI = "${GNOME_MIRROR}/glib/${SHRT_VER}/glib-${PV}.tar.xz \
            file://0001-Do-not-write-bindir-into-pkg-config-files.patch \
            file://0001-meson.build-do-not-hardcode-linux-as-the-host-system.patch \
            file://0001-meson-do-a-build-time-check-for-strlcpy-before-attem.patch \
+           file://0001-gdatetime.c-workaround-to-fix-test_month_names-1759.patch \
            file://glib-meson.cross \
            "
 
diff --git a/meta/recipes-core/glib-2.0/glib.inc b/meta/recipes-core/glib-2.0/glib.inc
index 574bb38..1e8fea2 100644
--- a/meta/recipes-core/glib-2.0/glib.inc
+++ b/meta/recipes-core/glib-2.0/glib.inc
@@ -152,6 +152,13 @@ RDEPENDS_${PN}-ptest += "\
             tzdata-europe \
             tzdata-posix \
             shared-mime-info \
+            ${PN}-locale-ja \
+            ${PN}-locale-fr \
+            ${PN}-locale-el \
+            ${PN}-locale-hr \
+            ${PN}-locale-lt \
+            ${PN}-locale-pl \
+            ${PN}-locale-ru \
            "
 
 RDEPENDS_${PN}-ptest_append_libc-glibc = "\
-- 
2.7.4

Re: [PATCH v2] glib-2.0: fix ptest failures

[Burton, Ross]

On Tue, 30 Apr 2019 at 08:45, <mingli.yu@windriver.com> wrote:
> From: Mingli Yu <Mingli.Yu@windriver.com>
>
> * Install 7 translation packages to fix 7 ptest
>   failures such as:
>   | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1477:test_non_utf8_printf: assertion failed (__p == ("10\346\234\210")): ("Oct" == "10\346\234\210")
>   | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1716:test_month_names: assertion failed (p_casefold == (o_casefold)): ("nov" == "stu")
>   | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1731:test_month_names: assertion failed (p_casefold == (o_casefold)): ("jul" == "liep.")

Would be better if we had a meta-package to install 'all
translations', but we don't.

> * Increase the timeout to 15 minutes as the default
>   timeout which is 5 minutes is not enough to fix below error:
>   | Executing: glib/actions.test
>   | Test timed out after 300 seconds
>   | cleaning up pid 13050
>   | FAIL: glib/actions.test (Child process killed by signal 9)

Looks like we're working around a test case that is going to fail
because it is timing out.  Let's investigate why it is taking so long
instead.

> * There are 240+ ptest cases for glib-2.0 and one
>   case is gdatetime.test. And there are 53 sub-cases
>   under gdatetime.test.
>
>   Among of the 53 sub-cases of gdatetime.test, there
>   is one sub-case named test_month_names.
>
>   For test_month_names case, there are also 64 cases
>   against it. Comment 2 cases of test_month_names
>   as a workaround to fix the below test_month_names
>   failure as the issue also exists in the latest version:
>   | GLib:ERROR:../glib-2.60.1/glib/tests
>   /gdatetime.c:1702:test_month_names: assertion failed (p_casefold == (o_casefold)): ("\316\261\317\205\316\263" == "\316\261\317\215\316\263")
>   | GLib:ERROR:../glib-2.60.1/glib/tests/gdatetime.c:1765:test_month_names: assertion failed (p_casefold == (o_casefold)): ("\320\274\320\260\321\217" == "\320\274\320\260\320\271")
>
>   Have reported the above issues to upstream as
>   https://gitlab.gnome.org/GNOME/glib/issues/1759
>   and as the failure even exists on the latest version 2.60.1
>   until now, so comment it as a workaround until the
>   upstream fixs it.

The test is failing and we don't know why, so it should continue to
fail.  We don't get to 100% pass rate by disabling every test that
fails.

Can you drop the timeout and datetime test removal, leaving just the
translation additions?

Ross

Re: [PATCH v2] glib-2.0: fix ptest failures

[Yu, Mingli]

On 2019年04月30日 21:16, Burton, Ross wrote:
> On Tue, 30 Apr 2019 at 08:45, <mingli.yu@windriver.com> wrote:
>> From: Mingli Yu <Mingli.Yu@windriver.com>
>>
>> * Install 7 translation packages to fix 7 ptest
>>    failures such as:
>>    | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1477:test_non_utf8_printf: assertion failed (__p == ("10\346\234\210")): ("Oct" == "10\346\234\210")
>>    | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1716:test_month_names: assertion failed (p_casefold == (o_casefold)): ("nov" == "stu")
>>    | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1731:test_month_names: assertion failed (p_casefold == (o_casefold)): ("jul" == "liep.")
>
> Would be better if we had a meta-package to install 'all
> translations', but we don't.

>
>> * Increase the timeout to 15 minutes as the default
>>    timeout which is 5 minutes is not enough to fix below error:
>>    | Executing: glib/actions.test
>>    | Test timed out after 300 seconds
>>    | cleaning up pid 13050
>>    | FAIL: glib/actions.test (Child process killed by signal 9)
>
> Looks like we're working around a test case that is going to fail
> because it is timing out.  Let's investigate why it is taking so long
> instead.

The sub-case named test_dbus_threaded[/actions/dbus/threaded] of 
glib/actions.test takes almost most of the time, the more details as below.

# time; /usr/libexec/installed-tests/glib/actions -p;time

user	0m58.87s

sys	0m16.89s

/actions/basic: OK

/actions/name: OK

/actions/simplegroup: OK

/actions/stateful: OK

/actions/default-activate: OK

/actions/entries: OK

/actions/parse-detailed: OK

/actions/property: OK

/actions/dbus/export: OK

/actions/dbus/threaded: OK

/actions/dbus/bug679509: OK

user	11m37.67
s
sys	2m20.77s



# time; /usr/libexec/installed-tests/glib/actions -p; time
user	0m0.11s
sys	0m0.05s
/actions/dbus/threaded: OK
user	6m32.04s
sys	0m55.54s


About the /actions/dbus/threaded,  check 
https://gitlab.gnome.org/GNOME/glib/blob/master/gio/tests/actions.c#L918 
for more details.


# gnome-desktop-testing-runner --help
Usage:
   gnome-desktop-testing-runner [OPTION?] [PREFIX...] - Run installed tests
[snip]
-t, --timeout=TIMEOUT       Cancel test after timeout seconds; defaults 
to 5 minutes
[snip]


If use gnome-desktop-testing-runner to run all the glib-2.0 test, we 
should guarantee all the tests have enough time to run and I don't think 
the below logic is a workaround, we should increase the timeout if use 
gnome-desktop-testing-runner.

-su glib2-test -c gnome-desktop-testing-runner glib
+su glib2-test -c "gnome-desktop-testing-runner glib -t 900"

Thanks,

>
>> * There are 240+ ptest cases for glib-2.0 and one
>>    case is gdatetime.test. And there are 53 sub-cases
>>    under gdatetime.test.
>>
>>    Among of the 53 sub-cases of gdatetime.test, there
>>    is one sub-case named test_month_names.
>>
>>    For test_month_names case, there are also 64 cases
>>    against it. Comment 2 cases of test_month_names
>>    as a workaround to fix the below test_month_names
>>    failure as the issue also exists in the latest version:
>>    | GLib:ERROR:../glib-2.60.1/glib/tests
>>    /gdatetime.c:1702:test_month_names: assertion failed (p_casefold == (o_casefold)): ("\316\261\317\205\316\263" == "\316\261\317\215\316\263")
>>    | GLib:ERROR:../glib-2.60.1/glib/tests/gdatetime.c:1765:test_month_names: assertion failed (p_casefold == (o_casefold)): ("\320\274\320\260\321\217" == "\320\274\320\260\320\271")
>>
>>    Have reported the above issues to upstream as
>>    https://gitlab.gnome.org/GNOME/glib/issues/1759
>>    and as the failure even exists on the latest version 2.60.1
>>    until now, so comment it as a workaround until the
>>    upstream fixs it.
>
> The test is failing and we don't know why, so it should continue to
> fail.  We don't get to 100% pass rate by disabling every test that
> fails.
>
> Can you drop the timeout and datetime test removal, leaving just the
> translation additions?
>
> Ross
>

Re: [PATCH v2] glib-2.0: fix ptest failures

[Adrian Bunk]

On Sun, May 05, 2019 at 02:09:49PM +0800, Yu, Mingli wrote:
> On 2019年04月30日 21:16, Burton, Ross wrote:
> > On Tue, 30 Apr 2019 at 08:45, <mingli.yu@windriver.com> wrote:
>...
> > > * Increase the timeout to 15 minutes as the default
> > >    timeout which is 5 minutes is not enough to fix below error:
> > >    | Executing: glib/actions.test
> > >    | Test timed out after 300 seconds
> > >    | cleaning up pid 13050
> > >    | FAIL: glib/actions.test (Child process killed by signal 9)
> > 
> > Looks like we're working around a test case that is going to fail
> > because it is timing out.  Let's investigate why it is taking so long
> > instead.
> 
> The sub-case named test_dbus_threaded[/actions/dbus/threaded] of
> glib/actions.test takes almost most of the time, the more details as below.
>...

On what hardware it this?

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: [PATCH v2] glib-2.0: fix ptest failures

[Yu, Mingli]

On 2019年05月06日 17:50, Adrian Bunk wrote:
> On Sun, May 05, 2019 at 02:09:49PM +0800, Yu, Mingli wrote:
>> On 2019年04月30日 21:16, Burton, Ross wrote:
>>> On Tue, 30 Apr 2019 at 08:45, <mingli.yu@windriver.com> wrote:
>> ...
>>>> * Increase the timeout to 15 minutes as the default
>>>>     timeout which is 5 minutes is not enough to fix below error:
>>>>     | Executing: glib/actions.test
>>>>     | Test timed out after 300 seconds
>>>>     | cleaning up pid 13050
>>>>     | FAIL: glib/actions.test (Child process killed by signal 9)
>>>
>>> Looks like we're working around a test case that is going to fail
>>> because it is timing out.  Let's investigate why it is taking so long
>>> instead.
>>
>> The sub-case named test_dbus_threaded[/actions/dbus/threaded] of
>> glib/actions.test takes almost most of the time, the more details as below.
>> ...
>
> On what hardware it this?


Hi Adrian,

I run the test in qemu and the resource info for the qemu target as below:

# time; /usr/libexec/installed-tests/glib/actions -p;time
user	0m0.11s
sys	0m0.03s
/actions/dbus/threaded: OK
user	6m21.03s
sys	0m45.91s


# cat /proc/cpuinfo
processor	: 0
vendor_id	: GenuineIntel
cpu family	: 6
model		: 15
model name	: Intel(R) Core(TM)2 Duo CPU     T7700  @ 2.40GHz
stepping	: 11
cpu MHz		: 3192.391
cache size	: 16384 KB
physical id	: 0
siblings	: 1
core id		: 0
cpu cores	: 1
apicid		: 0
initial apicid	: 0
fpu		: yes
fpu_exception	: yes
cpuid level	: 10
wp		: yes
flags		: fpu de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat 
pse36 clflush acpi mmx fxsr sse sse2 ss syscall nx lm constant_tsc rep_gi
bugs		: cpu_meltdown spectre_v1 spectre_v2 spec_store_bypass l1tf
bogomips	: 6384.78
clflush size	: 64
cache_alignment	: 64
address sizes	: 40 bits physical, 48 bits virtual
power management:


# cat /proc/meminfo
MemTotal:        3071748 kB
MemFree:         2899224 kB
MemAvailable:    2917468 kB
Buffers:            5044 kB
Cached:            49976 kB
SwapCached:            0 kB
Active:            94872 kB
Inactive:          35160 kB
Active(anon):      75484 kB
Inactive(anon):     8428 kB
Active(file):      19388 kB
Inactive(file):    26732 kB
Unevictable:           0 kB
Mlocked:               0 kB
SwapTotal:             0 kB
SwapFree:              0 kB
Dirty:                 0 kB
Writeback:             0 kB
AnonPages:         75008 kB
Mapped:            23808 kB
Shmem:              8904 kB
KReclaimable:      14436 kB
Slab:              27836 kB
SReclaimable:      14436 kB
SUnreclaim:        13400 kB
KernelStack:        1360 kB
PageTables:          976 kB
NFS_Unstable:          0 kB
Bounce:                0 kB
WritebackTmp:          0 kB
CommitLimit:     1535872 kB
Committed_AS:     181504 kB
VmallocTotal:   34359738367 kB
VmallocUsed:           0 kB
VmallocChunk:          0 kB
Percpu:              296 kB
DirectMap4k:       30576 kB
DirectMap2M:     3115008 kB

Thanks,

>
> cu
> Adrian
>

[PATCH v3] glib-2.0: fix ptest failures

[mingli.yu]

From: Mingli Yu <Mingli.Yu@windriver.com>

* Install 7 translation packages to fix 7 ptest
  failures such as:
  | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1477:test_non_utf8_printf: assertion failed (__p == ("10\346\234\210")): ("Oct" == "10\346\234\210")
  | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1716:test_month_names: assertion failed (p_casefold == (o_casefold)): ("nov" == "stu")
  | GLib:ERROR:../glib-2.58.3/glib/tests/gdatetime.c:1731:test_month_names: assertion failed (p_casefold == (o_casefold)): ("jul" == "liep.")

* Increase the timeout to 15 minutes as the default
  timeout which is 5 minutes is not enough to fix below error:
  | Executing: glib/actions.test
  | Test timed out after 300 seconds
  | cleaning up pid 13050
  | FAIL: glib/actions.test (Child process killed by signal 9)

Signed-off-by: Mingli Yu <Mingli.Yu@windriver.com>
---
 meta/recipes-core/glib-2.0/glib-2.0/run-ptest | 2 +-
 meta/recipes-core/glib-2.0/glib.inc           | 7 +++++++
 2 files changed, 8 insertions(+), 1 deletion(-)

diff --git a/meta/recipes-core/glib-2.0/glib-2.0/run-ptest b/meta/recipes-core/glib-2.0/glib-2.0/run-ptest
index 8f082d34f6..893125d772 100644
--- a/meta/recipes-core/glib-2.0/glib-2.0/run-ptest
+++ b/meta/recipes-core/glib-2.0/glib-2.0/run-ptest
@@ -2,5 +2,5 @@
 
 set -eux
 useradd glib2-test
-su glib2-test -c gnome-desktop-testing-runner glib
+su glib2-test -c "gnome-desktop-testing-runner glib -t 900"
 userdel glib2-test
diff --git a/meta/recipes-core/glib-2.0/glib.inc b/meta/recipes-core/glib-2.0/glib.inc
index 574bb38557..1e8fea2df1 100644
--- a/meta/recipes-core/glib-2.0/glib.inc
+++ b/meta/recipes-core/glib-2.0/glib.inc
@@ -152,6 +152,13 @@ RDEPENDS_${PN}-ptest += "\
             tzdata-europe \
             tzdata-posix \
             shared-mime-info \
+            ${PN}-locale-ja \
+            ${PN}-locale-fr \
+            ${PN}-locale-el \
+            ${PN}-locale-hr \
+            ${PN}-locale-lt \
+            ${PN}-locale-pl \
+            ${PN}-locale-ru \
            "
 
 RDEPENDS_${PN}-ptest_append_libc-glibc = "\
-- 
2.18.1

Re: [PATCH v3] glib-2.0: fix ptest failures

[Burton, Ross]

On Wed, 8 May 2019 at 04:06, <mingli.yu@windriver.com> wrote:
> * Increase the timeout to 15 minutes as the default
>   timeout which is 5 minutes is not enough to fix below error:
>   | Executing: glib/actions.test
>   | Test timed out after 300 seconds
>   | cleaning up pid 13050
>   | FAIL: glib/actions.test (Child process killed by signal 9)

My Debian machine runs that test in 20 seconds.

Maybe we're missing dependencies to get a working session dbus up?

Ross

Re: [PATCH v3] glib-2.0: fix ptest failures

[richard.purdie]

On Wed, 2019-05-08 at 12:54 +0100, Burton, Ross wrote:
> On Wed, 8 May 2019 at 04:06, <mingli.yu@windriver.com> wrote:
> > * Increase the timeout to 15 minutes as the default
> >   timeout which is 5 minutes is not enough to fix below error:
> >   | Executing: glib/actions.test
> >   | Test timed out after 300 seconds
> >   | cleaning up pid 13050
> >   | FAIL: glib/actions.test (Child process killed by signal 9)
> 
> My Debian machine runs that test in 20 seconds.
> 
> Maybe we're missing dependencies to get a working session dbus up?

That seems likely. I'd note that we see:

https://autobuilder.yocto.io/pub/non-release/20190502-12/testresults/testresult-report.txt


Recipe                       | Passed      | Failed   | Skipped   | Time(s)    
glib-2.0                     | 290         | 1        | 0         | 427

and the failing test is:

    ptestresult.glib-2.0.glib/gdatetime.test__Child_process_killed_by_signal_6_

so it would seem this is working on our autobuilder tests? That would
seem to back up Ross' comments.

Cheers,

Richard

Re: [PATCH v3] glib-2.0: fix ptest failures

[Yu, Mingli]

On 2019年05月08日 20:25, richard.purdie@linuxfoundation.org wrote:
> On Wed, 2019-05-08 at 12:54 +0100, Burton, Ross wrote:
>> On Wed, 8 May 2019 at 04:06, <mingli.yu@windriver.com> wrote:
>>> * Increase the timeout to 15 minutes as the default
>>>    timeout which is 5 minutes is not enough to fix below error:
>>>    | Executing: glib/actions.test
>>>    | Test timed out after 300 seconds
>>>    | cleaning up pid 13050
>>>    | FAIL: glib/actions.test (Child process killed by signal 9)
>>
>> My Debian machine runs that test in 20 seconds.
>>
>> Maybe we're missing dependencies to get a working session dbus up?
>
> That seems likely. I'd note that we see:
>
> https://autobuilder.yocto.io/pub/non-release/20190502-12/testresults/testresult-report.txt
>
>
> Recipe                       | Passed      | Failed   | Skipped   | Time(s)
> glib-2.0                     | 290         | 1        | 0         | 427
>
> and the failing test is:
>
>      ptestresult.glib-2.0.glib/gdatetime.test__Child_process_killed_by_signal_6_
>
> so it would seem this is working on our autobuilder tests? That would
> seem to back up Ross' comments.

Thanks Ross and RP's comments!

It's weird! Seems it succeeds in autobuilder, but I can reproduce the 
glib/actions.test failure steadily in my env as below:
# time; /usr/libexec/installed-tests/glib/actions -p;time
user	0m0.13s
sys	0m0.07s
/actions/dbus/threaded: OK
user	6m29.03s
sys	0m50.46s

Thanks,

>
> Cheers,
>
> Richard
>
>

Re: [PATCH v3] glib-2.0: fix ptest failures

[richard.purdie]

On Mon, 2019-05-13 at 17:25 +0800, Yu, Mingli wrote:
> 
> On 2019年05月08日 20:25, richard.purdie@linuxfoundation.org wrote:
> > On Wed, 2019-05-08 at 12:54 +0100, Burton, Ross wrote:
> > > On Wed, 8 May 2019 at 04:06, <mingli.yu@windriver.com> wrote:
> > > > * Increase the timeout to 15 minutes as the default
> > > >    timeout which is 5 minutes is not enough to fix below error:
> > > >    | Executing: glib/actions.test
> > > >    | Test timed out after 300 seconds
> > > >    | cleaning up pid 13050
> > > >    | FAIL: glib/actions.test (Child process killed by signal 9)
> > > 
> > > My Debian machine runs that test in 20 seconds.
> > > 
> > > Maybe we're missing dependencies to get a working session dbus
> > > up?
> > 
> > That seems likely. I'd note that we see:
> > 
> > https://autobuilder.yocto.io/pub/non-release/20190502-12/testresults/testresult-report.txt
> > 
> > 
> > Recipe                       | Passed      | Failed   | Skipped   |
> > Time(s)
> > glib-2.0                     | 290         | 1        | 0         |
> > 427
> > 
> > and the failing test is:
> > 
> >      ptestresult.glib-
> > 2.0.glib/gdatetime.test__Child_process_killed_by_signal_6_
> > 
> > so it would seem this is working on our autobuilder tests? That
> > would
> > seem to back up Ross' comments.
> 
> Thanks Ross and RP's comments!
> 
> It's weird! Seems it succeeds in autobuilder, but I can reproduce
> the 
> glib/actions.test failure steadily in my env as below:
> # time; /usr/libexec/installed-tests/glib/actions -p;time
> user	0m0.13s
> sys	0m0.07s
> /actions/dbus/threaded: OK
> user	6m29.03s
> sys	0m50.46s

I suspect its a dependency such as the dbus session Ross mentioned.
Which image are you running this test in? Have you tried testing in
something like a core-image-sato for comparison?

It would be good to narrow down the missing dependency.

Cheers,

Richard

Re: [PATCH v3] glib-2.0: fix ptest failures

[Yu, Mingli]

On 2019年05月13日 17:40, richard.purdie@linuxfoundation.org wrote:
> On Mon, 2019-05-13 at 17:25 +0800, Yu, Mingli wrote:
>>
>> On 2019年05月08日 20:25, richard.purdie@linuxfoundation.org wrote:
>>> On Wed, 2019-05-08 at 12:54 +0100, Burton, Ross wrote:
>>>> On Wed, 8 May 2019 at 04:06, <mingli.yu@windriver.com> wrote:
>>>>> * Increase the timeout to 15 minutes as the default
>>>>>     timeout which is 5 minutes is not enough to fix below error:
>>>>>     | Executing: glib/actions.test
>>>>>     | Test timed out after 300 seconds
>>>>>     | cleaning up pid 13050
>>>>>     | FAIL: glib/actions.test (Child process killed by signal 9)
>>>>
>>>> My Debian machine runs that test in 20 seconds.
>>>>
>>>> Maybe we're missing dependencies to get a working session dbus
>>>> up?
>>>
>>> That seems likely. I'd note that we see:
>>>
>>> https://autobuilder.yocto.io/pub/non-release/20190502-12/testresults/testresult-report.txt
>>>
>>>
>>> Recipe                       | Passed      | Failed   | Skipped   |
>>> Time(s)
>>> glib-2.0                     | 290         | 1        | 0         |
>>> 427
>>>
>>> and the failing test is:
>>>
>>>       ptestresult.glib-
>>> 2.0.glib/gdatetime.test__Child_process_killed_by_signal_6_
>>>
>>> so it would seem this is working on our autobuilder tests? That
>>> would
>>> seem to back up Ross' comments.
>>
>> Thanks Ross and RP's comments!
>>
>> It's weird! Seems it succeeds in autobuilder, but I can reproduce
>> the
>> glib/actions.test failure steadily in my env as below:
>> # time; /usr/libexec/installed-tests/glib/actions -p;time
>> user	0m0.13s
>> sys	0m0.07s
>> /actions/dbus/threaded: OK
>> user	6m29.03s
>> sys	0m50.46s
>
> I suspect its a dependency such as the dbus session Ross mentioned.
> Which image are you running this test in? Have you tried testing in
> something like a core-image-sato for comparison?

Use core-image-sato to test in my env as below and still 
/actions/dbus/threaded takes most of the time.

# time; /usr/libexec/installed-tests/glib/actions -p;time
user	0m0.20s
sys	0m0.09s
/actions/basic: OK
/actions/name: OK
/actions/simplegroup: OK
/actions/stateful: OK
/actions/default-activate: OK
/actions/entries: OK
/actions/parse-detailed: OK
/actions/property: OK
/actions/dbus/export: OK
/actions/dbus/threaded: OK
/actions/dbus/bug679509: OK
user	6m29.77s
sys	0m48.97s

Thanks,

>
> It would be good to narrow down the missing dependency.
>
> Cheers,
>
> Richard
>
>